Documentation

Python Documentation

Applies to both C++ and Python API. In Python use numpy arrays and matrices instead.

audio_to_novelty_curve

Arguments

• signal: wavefrom of audio signal
• sr: sampling rate of the audio (Hz)
• window_length: window length for STFT (in samples)
• hop_length: stepsize for the STFT
• compression_c: constant for log compression
• log_compression: enable/disable log compression
• resample_feature_rate: feature rate of the resulting novelty curve (resampled, independent of stepsize)

Returns
Tuple(novelty_curve, feature_rate)

Description Computes a novelty curve (onset detection function) for the input audio signal. This implementation is a variant of the widely used spectral flux method with additional bandwise processing and a logarithmic intensity compression. This particularly addresses music with weak onset information (e.g., exhibiting string instruments.)

novelty_curve_to_tempogram_dft

Arguments

• novelty_curve: a novelty curve indicating note onset positions
• bpm: vector containing BPM values to compute
• feature_rate: feature rate of the novelty curve (Hz). This needs to be set to allow for setting other parameters in seconds!
• tempo_window: Analysis window length in seconds
• hop_length: window hop length in frames (of novelty curve)

Returns
Tuple(tempogram, time vector, bpm vector)

Description Computes a complex valued fourier tempogram for a given novelty curve indicating note onset candidates in the form of peaks. This implementation provides parameters for chosing fourier coefficients in a frequency range corresponding to musically meaningful tempo values in bpm.

normalize_feature

Arguments

• feature: matrix
• p
• threshold

Returns
Normalized feature matrix

Description
Normalizes a feature sequence according to the l^p norm If the norm falls below threshold for a feature vector, then the normalized feature vector is set to be the unit vector.

compute_fourier_coefficients

Arguments

• s time domain signal
• window vector containing window function
• n_overlap overlap given in samples
• f vector of frequencies values of fourier coefficients, in Hz
• sr sampling rate of signal s in Hz

Returns
Tuple(complex fourier coefficients, frequency vector, time vector)

Description
Function that calculates a fourier coefficient with frequency f.

stft

Arguments

• signal: wavefrom of audio signal
• sr: sample rate
• window
• coefficient_range
• n_fft: window length
• hop_length

Returns
Tuple(spectrogram, frequency vector, time vector)

Description
Computes aspectrogram using a STFT (short-time fourier transform)